INDEX

Explanations

locations or names of places

oai_token-act-pair · gpt-3.5-turbo

names of specific geographic locations and individuals

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.1.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

innacle

-0.66

 stricken

-0.64

adobe

-0.64

estern

-0.63

ilts

-0.62

tre

-0.61

Tre

-0.60

RH

-0.60

calling

-0.59

ATURE

-0.59

POSITIVE LOGITS

 Hyder

1.35

 Pradesh

0.99

alore

0.94

abad

0.92

beans

0.82

 mosqu

0.81

sey

0.79

 lapt

0.79

Rao

0.78

omet

0.78

Activations Density 0.005%

locations or names of places

names of specific geographic locations and individuals

No Comments

No Known Activations